@@ -17,84 +17,94 @@ polynomial.
1717 ------------------------------------
1818
1919Well, this should be quite automatic. However, you need lapack and
20- blas in order to build the phmc_tm executable. Make sure that you have
20+ blas in order to build the executable. Make sure that you have
2121both available and configure/make the hmc code accordingly. If
22- successful, you will find the executable phmc_tm in your build
22+ successful, you will find the executable hmc_tm in your build
2323directory.
2424
2525 1. Setup the "hmc.input" file
2626 ------------------------------------
2727
28- Set the parameters for your desired simulation (only 2+1+1 specific parameters are
29- discussed here) :
30- hmc.input:
31-
32- # hopping parameter
33- # set this to kappa_critical for maximum twist
34- kappa = 0.1234
35-
36- # twisted mass, set this parameter to
37- # 2*kappa*mu_desired
38- # where mu_desired is the quark mass you want to simulate
39- # in the LIGHT quark sector
40- mu = 0.00005
41-
42- # twisted mass for the heavy sector
43- # set this to the mean value of the mass of the HEAVY quarks
44- # you want to simulate e.g. M_s=0.02 M_c=0.2 ==>
45- PhmcMubar = 0.11
46-
47- # split of the mass in the HEAVY quark sector
48- # set this to (M_c-M_s)/2
49- # such that you get finally:
50- # M_s=PhmcMubar - PhmcEpsbar
51- # M_c=PhmcMubar + PhmcEpsbar
52- PhmcEpsbar = 0.09
53-
54- # As we want to simulate "2+1+1" we have to say this to the program
55- PhmcNoFlavours = 2+1+1
56-
57- # If you want to calculate the eigenvalues every n'th trajectory
58- # then set this parameter to n if you want no eigenvalues set this to 0
59- # during thermalization you should set this to 1 or 2 to follow the evolution
60- # of smallest and largest eigenvalue to adjust the approximation interval
61- # of the polynomial (see below)
62- PhmcRecEVInterval = 1
63-
64- # Setting the polynomial parameters:
65- # ---------------------------------
66- # you should have a first estimate for the smallest and largest eigenvalue
67- # of the HEAVY quark operator. as an estimate you can take
68- # PhmcStildeMin = 2*(PhmcMubar^2-PhmcEpsbar^2) and
69- # PhmcStildeMax = 1/2/sqrt(PhmcMubar^2 + PhmcEpsbar^2) (taken from
70- # [1] )
71- # set the following parameter to your estimate of the smallest eigenvalue
72- PhmcStildeMin = 0.008
73-
74- # and this to an upper bound of all eigenvalues
75- #
76- PhmcStildeMax = 3.6
77-
78- # an estimate for the error of an approximation of 1/sqrt(x) by a chebycheff
79- # polynomial is given by (found in [2]):
80- # delta = 2*((1-sqrt(ratio))/(1+sqrt(ratio)))^(n+1)
81- # where ratio = PhmcStildeMin/PhmcStildeMax
82- # and "n" is the degree of the chebycheff polynomial
83- # e.g.: if you take the values from above and want an delta of 0.0001
84- # you have to set the degree to 104:
85- PhmcDegreeOfP = 104
86-
87- # To make the calculation of the hamiltonian more precise the program uses
88- # a second polynomial Ptilde.
89- # with the following parameter you can adjust the
90- # precision of the whole approximation of 1/sqrt(x)= Ptilde(x)*P(x)*(1+Rtilde(x))
91- # with |Rtilde(x)| ~ PhmcPrecisionPtilde (see also [1])
92-
93- PhmcPrecisionPtilde = 1.0e-9
94-
95- # end of hmc.input
96-
97- Now you should have a proper hmc.input file
28+ Set the parameters for your desired simulation (only 1+1 specific parameters are
29+ discussed here) :
30+ hmc.input:
31+
32+ # hopping parameter
33+ # set this to kappa_critical for maximum twist
34+ kappa = 0.1234
35+
36+ # twisted mass, set this parameter to
37+ # 2*kappa*mu_desired
38+ # where mu_desired is the quark mass you want to simulate
39+ # in the LIGHT quark sector
40+ mu = 0.00005
41+
42+ # twisted mass for the heavy sector
43+ # set this to the mean value of the mass of the HEAVY quarks
44+ # you want to simulate e.g. M_s=0.02 M_c=0.2 ==>
45+ 2KappaMubar = 0.11
46+
47+ # split of the mass in the HEAVY quark sector
48+ # set this to (M_c-M_s)/2
49+ # such that you get finally:
50+ # M_s=PhmcMubar - PhmcEpsbar
51+ # M_c=PhmcMubar + PhmcEpsbar
52+ 2KappaEpsbar = 0.09
53+
54+ # Setting the polynomial parameters:
55+ # create a NDPOLY monomial first (see docu)
56+ BeginMonomial NDPOLY
57+
58+ # If you want to calculate the eigenvalues every n'th trajectory
59+ # then set this parameter to n if you want no eigenvalues set this to 0
60+ # during thermalization you should set this to 1 or 2 to follow the evolution
61+ # of smallest and largest eigenvalue to adjust the approximation interval
62+ # of the polynomial (see below)
63+ ComputeEVFreq = 1
64+
65+ # ---------------------------------
66+ # you should have a first estimate for the smallest and largest eigenvalue
67+ # of the HEAVY quark operator. as an estimate you can take
68+ # StildeMin = 2*(PhmcMubar^2-PhmcEpsbar^2) and
69+ # StildeMax = 1/2/sqrt(PhmcMubar^2 + PhmcEpsbar^2) (taken from
70+ # [1] )
71+ # set the following parameter to your estimate of the smallest eigenvalue
72+ # or a little below
73+ #
74+ StildeMin = 0.008
75+
76+ # and this to an upper bound of all eigenvalues
77+ #
78+ StildeMax = 3.6
79+
80+ # an estimate for the error of an approximation of 1/sqrt(x) by a chebycheff
81+ # polynomial is given by (found in [2]):
82+ # delta = 2*((1-sqrt(ratio))/(1+sqrt(ratio)))^(n+1)
83+ # where ratio = PhmcStildeMin/PhmcStildeMax
84+ # and "n" is the degree of the chebycheff polynomial
85+ # e.g.: if you take the values from above and want an delta of 0.0001
86+ # you have to set the degree to 104:
87+ DegreeOfMDPolynomial = 104
88+
89+ # To make the calculation of the hamiltonian more precise the program uses
90+ # a second polynomial Ptilde.
91+ # with the following parameter you can adjust the
92+ # precision of the whole approximation of 1/sqrt(x)= Ptilde(x)*P(x)*(1+Rtilde(x))
93+ # with |Rtilde(x)| ~ PhmcPrecisionPtilde (see also [1])
94+
95+ PrecisionPtilde = 1.0e-9
96+
97+ # and you have to specify on which timescale to integrate the 1+1 part
98+ Timescale = 1
99+
100+ EndMonomial
101+ # you will need other monomials and an integrator, see docu
102+ # end of hmc.input
103+
104+ Now you should have a proper hmc.input file. For a complete example
105+ see the file
106+ hmc/sample-input/sample-hmc2.input
107+ and the comments therein.
98108
991092. Creating a polynomial
100110-----------------------------------
@@ -146,11 +156,11 @@ During the thermalization you should keep track of the lowest and largest
146156 eigenvalue. this can be done by filtering the job output file.
147157cat job.12345.out|grep -A6 LAMBDA
148158. Alternating you get the (actually) four smallest and four largest
149- eigenvalues normalized to the value of PhmcStildeMax you set in
159+ eigenvalues normalized to the value of StildeMax you set in
150160 hmc.input file. To get the unscaled eigenvalue you have to multiply the
151- values by PhmcStildeMax .
161+ values by StildeMax .
152162If the eigenvalues you measured are outside the approximation interval
153- [PhmcStildeMin,PhmcStildeMax ] you have to re-adjust these values (in hmc.input) and
163+ [StildeMin, StildeMax ] you have to re-adjust these values (in hmc.input) and
154164also regenerate the files
155165
156166 Square_root_BR_roots.dat
0 commit comments